High-Density Fiberboard (HDF) has been produced globally in vast quantities over the last few years. Rejects in production are common due to high variability in board properties. This work statistically analyzes a HDF production plant, with the aim of finding the major sources of variation. Process data from all processing steps were used; and also data from the raw material (wood species, acidic groups measured by ion chromatography, hemicelluloses, extractives determined through methanolysis, pH, buffer capacities), and subjective variables such as process performance or formaldehyde perceptibility. As response variables, the board properties Internal Bond Strength, Surface Soundness (, Modulus of Rupture, Modulus of Elasticity and Thickness Swelling 24 h , further press factor and resin fraction were analyzed. Overall, the dataset consisted of 251 observations and 245 variables. Production lag times were considered. Partial least squares regression (PLSR) was used to create 45 models. The main key sources of variation were determined by (1) the frequency in which variables occur in models and by (2) weighting the regression coefficients according to the technological relevance of the board properties. The models show that board properties were influenced by the raw material variables to an average of 21%, while the remaining variance is absorbed by the process variables. Furthermore, the appropriateness of (multivariate) control charting as a tool of Statistical Process Control (SPC) is shown with the data.